AlterLab MCP Server is a Model Context Protocol (MCP) server that provides web scraping, structured data extraction, and screenshot capabilities for Claude, Cursor, Windsurf, and other AI agents. Its core goal is to replace AI's built-in WebFetch tool and open-source browser MCP servers, addressing their inability to bypass Cloudflare, render JavaScript SPAs, and extract structured data.
In terms of main capabilities, AlterLab uses a multi-tier scraping architecture that automatically selects the cheapest method capable of retrieving each URL. The tiers include: Curl ($0.0002/request, suitable for static pages, RSS, public APIs), HTTP ($0.0003/request, TLS fingerprint rotation), Stealth ($0.0005/request, browser disguise, handles Cloudflare/DataDome), Light JS ($0.0007/request, lightweight JS extraction), and Browser ($0.001/request, full headless Chromium, handles JavaScript-heavy SPAs). Auto mode starts from the first tier and only escalates when blocked. Scraping returns markdown by default, and also supports text, HTML, or JSON.
Structured data extraction supports preset configurations (products, articles, jobs, FAQs, recipes, events) or custom JSON Schema, returning clean JSON. The screenshot tool can take full-page screenshots of any URL and return PNG directly in the conversation. The cost estimation tool can show the estimated tier, per-request cost, and confidence before execution; balance inquiry requires no parameters. Session management tools support creating, listing, validating, and deleting stored sessions. Sessions contain cookies for specific domains, used to scrape content behind login walls (such as order history, account dashboards, members-only pricing). It supports two authentication methods: inline cookies and stored sessions, and emphasizes residential proxy rotation and automatic retries.
In terms of usage, it supports one-line command installation and setting the API key via the environment variable ALTERLAB_API_KEY. It can be integrated into Claude Desktop/Claude Code (~/.claude.json), Cursor (.cursor/mcp.json), Windsurf (~/.claudedeium/windsurf/mcp_config.json), and also supports Smithery installation. You need to register on the official website to obtain an API key.
Applicable scenarios include: research and analysis (scrape search results and summarize, extract product prices, compare competitor pricing pages), code generation (scrape API documentation to generate TypeScript types, extract color schemes to generate Tailwind configurations, screenshot designs and replicate them with React), content and SEO (analyze blog posts, extract FAQ datasets, compare landing pages), monitoring and alerts (check inventory, price drop alerts, monitor job postings), and AI agent workflows such as research agents, data pipeline agents, competitive intelligence, content generation, lead enrichment, and compliance monitoring.
Compatible with any MCP-supporting client, including Claude Desktop/Code, Cursor, Windsurf, Cline, GitHub Copilot, custom agents (Python/TypeScript/Go MCP SDK), and n8n workflow automation.
In terms of limitations, free-tier accounts have rate limits, which are removed after adding balance; automatic retries with exponential backoff handle 429 errors. Batch scraping supports up to 100 URLs per run. Pricing is pay-as-you-go, with no subscription and no monthly minimum. $1 can support about 5,000 lightweight scrapes (depending on the target website). Pricing for other optional add-ons is not currently provided.
